claim
active
claim:attentional-circuits-employ-many-computations-and-heuristics-in-parallel-even-in-simple-contextsAttentional circuits employ many computations and heuristics in parallel, even in simple contexts
New mechanism insight.
Source paper
extracted_fromNeighborhood — ranked by edge-count
Papers (1)
paper
Findings (2)
finding
- QK attribution of induction-relevant heads.
- In the Aunt Mary Sue prompt, QK attributions matrix has rank at least 2 with no strong cross termssupportsParallel independent QK interactions.
Questions (1)
question
- Head polysemanticity.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Key decomposition enabling separate analysis of where attention goes and what it does
- Claim supported by VPD's recovery of cross-head attention subcomponents, noted in footnote.
- Lack of metacognitive insight into arithmetic.
- Identification of algorithms implemented in attention layers, distributed across attention headsfinding0.783VPD successfully recovered interpretable attention algorithms (previous-token behavior, syntax-boundary routing) in weight space without requiring manual decomposition across heads.
- Long-standing bottleneck in mechanistic interpretability that VPD addresses by working natively on attention weight matrices.
- Process using Q, K, V to compute a heat map over K and weighted sum of V.
- Forward-looking hypothesis positioned as a conclusion and future direction of the paper
- Will circuit discovery provide insights into arbitrary tasks, or only where tasks can be crisply defined?question0.768Circuit discovery scope.